Original Articles

Rational Truncation of an RNA Aptamerto Prostate-Specific Membrane Antigen Using

Computational Structural Modeling

William M. Rockey,1 Frank J. Hernandez,2,* Sheng-You Huang,3–6,* Song Cao,3,4,6,* Craig A. Howell,2

Gregory S. Thomas,7 Xiu Ying Liu,2 Natalia Lapteva,8 David M. Spencer,8 James O. McNamara II,2

Xiaoqin Zou,3–6 Shi-Jie Chen,3,4,6 and Paloma H. Giangrande1,2,7

RNA aptamers represent an emerging class of pharmaceuticals with great potential for targeted cancer diag-nostics and therapy. Several RNA aptamers that bind cancer cell-surface antigens with high affinity and spec-ificity have been described. However, their clinical potential has yet to be realized. A significant obstacle to theclinical adoption of RNA aptamers is the high cost of manufacturing long RNA sequences through chemicalsynthesis. Therapeutic aptamers are often truncated postselection by using a trial-and-error process, which istime consuming and inefficient. Here, we used a ‘‘rational truncation’’ approach guided by RNA structuralprediction and protein/RNA docking algorithms that enabled us to substantially truncateA9, an RNA aptamerto prostate-specific membrane antigen (PSMA),with great potential for targeted therapeutics. This truncatedPSMA aptamer (A9L; 41mer) retains binding activity, functionality, and is amenable to large-scale chemicalsynthesis for future clinical applications. In addition, the modeled RNA tertiary structure and protein/RNAdocking predictions revealed key nucleotides within the aptamer critical for binding to PSMA and inhibitingits enzymatic activity. Finally, this work highlights the utility of existing RNA structural prediction and pro-tein docking techniques that may be generally applicable to developing RNA aptamers optimized for thera-peutic use.

Introduction

RNA aptamers are synthetic, single-stranded oligonu-cleotide ligands typically 30 to 70 bases in length thatadopt complex 3-dimensional (3D) conformations to bindtargets with high affinity and specificity (Dassie et al., 2009;Keefe et al., 2010). The targets of RNA aptamers include smallmolecules, peptides, proteins (secreted factors, intracellularproteins, and membrane receptors), and even whole cells(Dassie et al., 2009; Keefe et al., 2010). High-affinity RNAaptamers for specific targets can be derived from combina-torial RNA sequence libraries (with complexities of *1014) byan iterative selection process termed SELEX (SystematicEvolution of Ligands by EXponential Enrichment) (Ellingtonand Szostak, 1990; Jellinek et al., 1995). To enable the use of

RNA aptamers for in vivo applications, modified nucleotides[eg, 2’-fluoropyrimidines (Ruckman et al., 1998; Bieseckeret al., 1999; Rusconi et al., 2002), 2’-amino pyrimidines (Linet al., 1994; Jellinek et al., 1995), or 2’-O-methyl ribose purinesand pyrimidines (Burmeister et al., 2005, 2006)] are usuallyincorporated during the selection process or postselectionduring chemical synthesis (Huang et al., 1997; Padilla andSousa, 1999).

The affinities and specificities of RNA aptamers for theirtargets are comparable to those of antibodies for their anti-gens. Similar to antibodies, RNA aptamers can be used fortargeted diagnostics and therapeutics. At the bench, RNAaptamers have been successfully used as inhibitors of theirtargets (Thiel and Giangrande, 2009) as well as to deliverchemotherapeutic agents (Bagalkot et al., 2006; Dhar et al.,

Departments of 1Radiation Oncology and 2Internal Medicine, University of Iowa, Iowa City, Iowa.Departments of 3Physics and Astronomy and 4Biochemistry, University of Missouri, Columbia, Missouri.5Dalton Cardiovascular Research Center, University of Missouri, Columbia, Missouri.6Informatics Institute, University of Missouri, Columbia, Missouri.7Molecular and Cellular Biology Program, University of Iowa, Iowa City, Iowa.8Department of Pathology and Immunology, Baylor College of Medicine, Houston, Texas.*These three authors contributed equally to this work.

NUCLEIC ACID THERAPEUTICSVolume 21, Number 5, 2011ª Mary Ann Liebert, Inc.DOI: 10.1089/nat.2011.0313

299

2008; Gu et al., 2008; Cao et al., 2009), nanoparticles (Far-okhzad et al., 2004), radionuclides (Hicke et al., 2006), andsiRNAs (Chu et al., 2006; McNamara et al., 2006; Zhou et al.,2008, 2009; Dassie et al., 2009; Pastor et al., 2010) to specific celltypes in culture and in vivo. Several RNA aptamers are cur-rently undergoing clinical trials (Biesecker et al., 1999; Dykeet al., 2006; Gilbert et al., 2007; Cosmi, 2009; Buff et al., 2010;Eikelboom et al., 2010; Mongelard and Bouvet, 2010) and one,Pegaptanib, was approved for therapeutic use in age-relatedmacular degeneration by the U.S. Food and Drug Adminis-tration in 2004 (Gragoudas et al., 2004; Chakravarthy et al.,2006; Ng and Adamis, 2006). As targeted therapeutic agents,RNA aptamers have several advantages compared with an-tibodies, such as smaller size, better tissue penetration, ease ofchemical synthesis/modification, and the lack of immunestimulation. Further, from the standpoint of pharmaceuticalmanufacturing, RNA aptamers are not classified as biologicalagents, thus easing regulatory approval.

Despite these advantages, a current obstacle to deliveringRNA aptamer technology to the clinic cost effectively is theability to chemically synthesize long RNAs ( > 60 nucleotides)in large-scale quantities (Reese, 2005). Aptamer production isbased on solid-phase phosphoroamidite chemistry via anautomated process used for small-scale oligonucleotide syn-thesis. This process is highly reproducible, thus allowing shortsynthetic RNA aptamers (15–50 nucleotides in length) to bepurified to a high degree of purity/stability and syntheticyield. However, RNA aptamers of long length remain difficultto synthesize under these conditions. Although the efficiencyof the manufacturing process for synthetic oligonucleotidescontinues to improve, perhaps the simplest way to ensurehigh synthetic yield is to decrease the length of the oligonu-cleotide sequence to be synthesized. One potential solution tothis problem is the identification of shorter RNA aptamersequences through the use of short RNA SELEX libraries ( < 50nucleotides in length). However, the downside to this ap-proach is a reduction in the sequence complexity of the overallRNA aptamer library that could compromise the identifica-tion of optimal sequences (Sassanfar and Szostak, 1993).

Several approaches have been described for reducing thelength of long RNA aptamers to minimal functional se-quences. These approaches often require significant experi-mental efforts (Burgstaller et al., 1995; Green et al., 1995;Katilius et al., 2007). Perhaps the most common method fortruncating RNA aptamers postselection is a trial-and-errorapproach that is often time consuming and arduous. A no-table example of this has been the truncation of RNA apta-mers that bind to prostate-specific membrane antigen (PSMA)(Lupold et al., 2002). The trial-and-error approach was suc-cessfully used by Lupold and colleagues to truncate one of 2nuclease-resistant RNA aptamers (A9 and A10) that had beenselected to inhibit PSMA enzymatic activity (Lupold et al.,2002). By consecutively removing 5 bases from the 3¢-terminus,the authors were able to truncate the A10 RNA aptamer from71 to 56 nucleotides (A10-3) while retaining functionality(ability to inhibit PSMA enzymatic activity) and ability to bein vitro transcribed by using a T7 RNA polymerase. However,when a similar truncation approach was applied to the A9aptamer in this study, the aptamer was rendered inactive.

Given the therapeutic potential of the PSMA RNA apta-mers for applications including inhibition of PSMA’s pro-carcinogenic properties (Silver et al., 1997; Lapidus et al., 2000;

Colombatti et al., 2009; Yao et al., 2010) and delivery of smallmolecule drugs/toxins (Bagalkot et al., 2006; Dhar et al., 2008,2011), therapeutic siRNAs (McNamara et al., 2006; Dassieet al., 2009; Pastor et al., 2010), and nanoparticles (Farokhzadet al., 2004) to prostate cancer cells, further optimization tofacilitate large-scale chemical synthesis of these RNAs iscompelling. Toward this end, we have employed computa-tional RNA structural modeling and RNA/protein dockingmodels to guide the truncation of the A9 PSMA RNA apta-mer. This analysis resulted in a truncated derivative of the A9aptamer (A9L, 41mer), which, due to its reduced length, isnow amenable to large-scale chemical synthesis. Importantly,A9L retains PSMA binding activity/specificity and function-ality. Specifically, we show that A9L inhibits PSMA’s enzy-matic activity and, when directly applied to cells expressingPSMA, is effectively internalized.

In summary, these studies demonstrate the utility of com-putational RNA secondary and tertiary structure models forguiding/enabling truncations of RNA aptamers while re-taining their function. Further, these studies have resulted inversions of the PSMA A9 aptamer that due to their shortersequence length are now amenable to large-scale chemicalsynthesis for therapeutic applications.

Materials and Methods

DNA templates and primers for generating the duplexDNA used for transcription of the RNA aptamers

A9a aptamer: DNA Template: 5¢-GGGAGGACGATGCGGACCGAAAAAGACCTGACTTCTATACTAAGTCTACGTTCCCAGACGACTCCC -3¢

5¢ primer: 5¢-TAATACGACTCACTATAGGGAGGACGATGCGGA-3¢

3¢ primer: 5¢-GGGAGTCGTCTGGGAA-3¢

A9baptamer: DNA Template: 5¢-GGGACGATGCGGACCGAAAAAGACCTGACTTCTATACTAAGTCTACGTTCCCAGACGCCC-3¢

5¢ primer: 5¢-TAATACGACTCACTATAGGGACGATGCGGACCG-3¢

3¢ primer: 5¢-GGGCGTCTGGGAACGT-3¢

A9c aptamer: DNA Template: 5¢-GGGATGCGGACCGAAAAAGACCTGACTTCTATACTAAGTCTACGTTCCCAGACCC-3¢

5¢ primer: 5¢-TAATACGACTCACTATAGGGATGCGGACCGAAA-3¢

3¢ primer: 5¢-GGGTCTGGGAACGTAG-3¢

A9daptamer: DNA Template: 5¢-GGGACGATGCGGACCGAAAAAGACCTGACTTCTATACTAAGTCTACGTTCCCAGACGACCC-3¢

5¢ primer: 5¢-TAATACGACTCACTATAGGGACGATGCGGACCG-3¢

3¢ primer: 5¢-GGGTCGTCTGGGAACG-3¢

A9eaptamer: DNA Template: 5¢-GGGCGGACCGAAAAAGACCTGACTTCTATACTAAGTCTACGTTCCCACC-3¢

5¢ primer: 5¢-TAATACGACTCACTATAGGGCGGACCGAAAAAG-3¢

3¢ primer: 5¢-GGTGGGAACGTAGACT-3¢

A9faptamer: DNA Template: 5¢-GGGCGGACCGAAAAAGACCTGACTTCTATACTAAGTCTACGTTCCCAG CCC-3¢

300 ROCKEY ET AL.

5¢ primer: 5¢-TAATACGACTCACTATAGGGCGGACCGAAAAAG-3¢

3¢ primer: 5¢-GGGCTGGGAACGTAGA-3¢A9g aptamer: DNA Template: 5¢-GGGACCGAAAAAGAC

CTGACTTCTATACTAAGTCTACGTTCCC-3¢5¢ primer: 5¢-TAATACGACTCACTATAGGGACCGAAAA

AGACC -3¢3¢ primer: 5¢-GGGAACGTAGACTTAG-3¢

Chemically synthesized double-stranded DNAtemplates used for transcription of the RNA aptamers

A9g aptamer: Sense: 5¢-TAATACGACTCACTATAGGGACCGAAAAAGACCTGACTTCTATACTAAGTCTAC GTTCCC-3¢

Antisense: 5¢-GGGAACGTAGACTTAGTATAGAAGTCAGGTCTTTTTCGGTCCCTATAGTGA GTCGTATTA -3¢

A9h aptamer: Sense: 5¢-TAATACGACTCACTATAGGGGAAAAAGACCTGACTTCTATACTAAGTCTACCCC-3¢

Antisense: 5¢-GGGGTAGACTTAGTATAGAAGTCAGGTCTTTTTCCCCTATAGTGAGTCGTA TTA -3¢

A9i aptamer: Sense: 5¢-TAATACGACTCACTATAGGGCCTGACTTCTATACTAAGCCC-3¢

Antisense: 5¢-GGGCTTAGTATAGAAGTCAGGCCCTATAGTGAGTCGTATTA-3¢

A9j aptamer: Sense: 5¢-TAATACGACTCACTATAGGGACCGAAAAAGACCTAGTCTACGTTCCC-3¢

Antisense: 5¢-GGGAACGTAGACTAGGTCTTTTTCGGTCCCTATAGTGAGTCGTATTA-3¢

A9k aptamer: Sense: 5¢-TAATACGACTCACTATAGGGACCGAAAAATACGTTCCC-3¢

Antisense: 5¢-GGGAACGTATTTTTCGGTCCCTATAGTGAGTCGTATTA-3¢

A9L aptamer: Sense: 5¢-TAATACGACTCACTATAGGGCCGAAAAAGACCTGACTTCTATACTAAGTCTACG TCCC-3¢

Antisense: 5¢-GGGACGTAGACTTAGTATAGAAGTCAGGTCTTTTTCGGCCCTATAGTGAGT CGTATTA-3¢

A9g.1 aptamer: Sense: 5¢-TAATACGACTCACTATAGGGACCGAAAAAGGCCTGACTTCTATACTAAGCCTACGTTCCC-3¢

Antisense: 5¢-GGGAACGTAGGCTTAGTATAGAAGTCAGGCCTTTTTCGGTCCCTATAGTGAGTCGTATTA-3¢

A9g.2 aptamer:Sense:5¢-TAATACGACTCACTATAGGG AC CGAAAAAGCCCTGACTTCTATACTAAGGCTAC GTT CCC-3¢

Antisense: 5¢-GGGAACGTAGCCTTAGTATAGAAGTCAGGGCTTTTTCGGTCCCTATAGTGAGTCGTATTA-3¢

A9g.3 aptamer: Sense: 5¢-TAATACGACTCACTATAGGGACCGAAAAAGACCTGACTTCTATACTAAGTCTACGGTCCC-3¢

Antisense: 5¢-GGGACCGTAGACTTAGTATAGAAGTCAGGTCTTTTTCGGTCCCTATAGTGAGTCGTATTA-3¢

A9g.4 aptamer: Sense: 5¢-TAATACGACTCACTATAGGGACCGAAAAAGACCTGACTTCTATACTAAGTCTTC GTTCCC-3¢

Antisense: 5¢-GGGAACGAAGACTTAGTATAGAAGTCAGGTCTTTTTCGGTCCCTATAGTGAGTCGTATTA -3¢

A9g.5 aptamer: Sense: 5¢-TAATACGACTCACTATAGGGACCGAAAAAGACCTGACTTCTATACTAGGTCTAC GTTCCC-3¢

Antisense: 5¢-GGGAACGTAGACCTAGTATAGAAGTCAGGTCTTTTTCGGTCCCTATAGTGAGTCGTATTA-3¢

A9g.6 aptamer: Sense: 5¢-TAATACGACTCACTATAGGGACCGAAAAAGACCTGGCTTCTATACTAAGTCTAC GTTCCC-3¢

Antisense: 5¢-GGGAACGTAGACTTAGTATAGAAGCCAGGTCTTTTTCGGTCCCTATAGTGAGTCGTATTA-3¢

A9g.7 aptamer: Sense: 5¢-TAATACGACTCACTATAGGGACCGAAAAAGACCTGACTTCTATACTAAGTCTAC GATCCC-3¢

Antisense: 5¢-GGGATCGTAGACTTAGTATAGAAGTCAGGTCTTTTTCGGTCCCTATAGTGA GTCGTATTA-3¢

A9g.8 aptamer: Sense: 5¢-TAATACGACTCACTATAGGGACCGAAAAAGACCTGACTTCTATACTAAGTCTAC GCTCCC-3¢

Antisense: 5¢-GGGAGCGTAGACTTAGTATAGAAGTCAGGTCTTTTTCGGTCCCTATAGTGAGTCGTATTA-3¢

RNA truncations

To generate the A9 truncations, the sequence of full-lengthA9 as previously reported (Lupold et al., 2002) (5¢-GGGAGGACGAUGCGGACCGAAAAAGACCUGACUUCUAUACUAAGUCUACGUUCCCAGACGACUCGCCCGA-3¢) wasloaded into the program RNAStructure 4.6 (Mathews, 2006;Mathews et al., 2007). Using a computer-guided ‘‘rationaltruncation’’ approach, bases were removed from the 5¢ and 3¢ends such that the predicted secondary structure of the re-maining oligonucleotide was as similar as possible to that offull-length A9. Where necessary, base changes were made atthe 5¢ and 3¢ ends to maintain a 5¢-GGG transcription startcodon and a complementary 3¢-CCC. To create the illustra-tions, the secondary structures were rendered with the pro-gram VARNA 3.7 (Darty et al., 2009).

RNA transcriptions

The RNA was transcribed as previously described (McNa-mara et al., 2006). Briefly, template DNAs and primers wereordered from Integrated DNA Technologies (IDT). Using theprimer and template sequences just described, the double-stranded DNA templates for transcription were generated aspreviously described (McNamara et al., 2006). DNA templateswere purified with Qiagen DNA purification columns (27106)and used in in vitro transcription reactions as described inMcNamara et al. (2006) to make individual RNA aptamers. AY639F mutant T7 RNA polymerase (Huang et al., 1997) wasused to incorporate 2’fluoro modified pyrimidines to renderthe RNAs resistant to nuclease degradation. The RNA from thetranscription was run on a denaturing 10% acrylamide/7Murea gel, visualized using UV shadowing. The RNA was ex-cised from the gel, eluted in 4 mL of TE buffer, washed twicewith 4 mL of TE buffer, and concentrated with an Amicon10,000 MW-cutoff spin filter (UFC801024).

As an alternative to amplifying the double-stranded DNAtemplates by polymerase chain reaction (PCR), the completesense and antisense strands of the RNA transcription templatewere ordered from IDT. To anneal the 2, each oligonucleotidestrand was added to 500 mL of PCR-grade H2O to a finalconcentration of 3mM per strand, heated to 72�C for 5 min-utes, and then allowed to cool to room temperature over 10

COMPUTER-GUIDED OPTIMIZATION OF AN RNA APTAMER 301

minutes. The resulting double-stranded DNA was used in anRNA transcription reaction as just described. The aptamersA9g, A9h, A9i, A9j, A9k, A9L, and all aptamer mutationswere transcribed from chemically synthesized double-stranded DNA templates in this fashion.

PSMA NAALADase activity assay

The PSMA NAALADase activity assay was modified froma previously published protocol (Xiao et al., 2000) and per-formed in a final reaction volume of 200 mL. Double-distilledH2O (ddH2O) was used in the reaction solutions. The RNAaptamers were refolded in binding buffer (20 mM HEPES,150 mM NaCl, and 2 mM CaCl2) at a concentration 1.667 timesthe final concentration desired in the activity assay (eg,333 nM for a final concentration of 200 nM). Refolding wasaccomplished by heating at 65�C for 10 minutes, followed bycooling to 37�C for 10 minutes. A volume of 120mL of refoldedRNA in binding buffer was added to an Eppendorf tube, wascombined with 40mL of 200 mM Tris buffer, pH 7.5, and 20 mL10 mM CoCl2 (final concentrations in the reaction 40 and1 mM, respectively). Cobalt (II) chloride was reported to be a‘‘stimulator of enzymatic activity’’ in the original NAALA-Dase assay protocol (Xiao et al., 2000). When this compoundwas omitted from the reaction, we observed increased non-specific RNA interactions. Two micrograms in 2mL of re-combinant human PSMA (4234-ZN-010) from R&D Systemswas diluted in 500mL of 50 mM pH 7.5 Tris buffer. Ten mi-croliters of the PSMA solution (40 ng PSMA) was added to thereaction mix, and the reaction was incubated for 5 minutes at37�C to promote RNA-PSMA interaction. For the experimentshown in Fig. 1A, recombinant, purified human PSMA wasobtained courtesy of Dr. David Spencer (Baylor College ofMedicine). In this experiment, 2.4mg of human recombinantPSMA protein in 10mL of 50 mM pH 7.5 Tris buffer was addedto each reaction. Ten microliters of a working solution con-taining 0.55mM NAAG in H2O having a specific activity of10 nCi/mL of [glutamate-3,4-3H]-NAAG from Perkin Elmer(NET1082250UC) was added to the reaction mixture. The re-action was allowed to proceed for 15 minutes, mixing once bypipetting at 7.5 minutes. To halt the reaction, an equal volume(200mL) of cold 0.1 M phosphate buffer (dibasic sodium phos-phate, Na2HPO4) was added to the reaction mixture.

AG 1-X8 formate resin (200–400 mesh) columns from Bio-Rad Laboratories (731-6221) were used to quantitate the [3H]-glutamate reaction product. Before use, the columns wereequilibrated with 5 mL of ddH2O. Half of the final reactionvolume (200mL) was added to a column. The columns wereeluted twice with 2 mL of 1 M formic acid. The first elution wasdiscarded, and the second 2 mL elution was added to 10 mL ofBio-Safe II scintillation fluid (Research Products InternationalCorp.). Activity was counted by using a Beckman-Coulterliquid scintillation counter, and was normalized to the amountof activity obtained in the reaction with no RNA added.

Filter binding assays

Filter binding assays were performed as previously de-scribed (Wong and Lohman, 1993). Briefly, aptamers were5¢-end labeled with 32P by using PNK. Labeled RNAs werediluted to 2000 cpms/mL in binding buffer, heated at 95�C for5 minutes to unfold the RNA, and allowed to refold at 37�Cfor 10 minutes. Five microliters of refolded labeled RNA was

added to each reaction. RNA was incubated for 5 minuteswith various concentrations (ranging from 1 to 1000 nM) ofpurified, recombinant human PSMA (4234-ZN-010) obtainedfrom R&D Systems at 37�C. The reaction mixture was spottedonto a sandwich of nitrocellulose (Protran BA 83, 0.2 mm poresize, 10 402 488; Whatman), nylon (Zeta-Probe BlottingMembranes, 162-0153; Bio-Rad Laboratories), and Whatman3MM chromatography paper (3130-6189) assembled in a dot-blot apparatus. Bound RNA was captured on the nitrocellu-lose filter, whereas unbound RNA was captured on the nylonfilter. The ratio of bound:unbound RNA was calculated byexposing the filters to a storage phosphor screen and imagingwith a phosphorimager.

Surface plasmon resonance (BIACore)binding measurements

Surface plasmon resonance (SPR) measurements werecarried out by using a BIACore 3000 device. 5¢-biotinylatedRNA was generated by transcription and gel purification asjust described, except that the transcription reactions werecarried out in the presence of 3 mM biotin-G (Custom orderfrom TriLink Biotechnologies: 5¢-(Biotin) (Spacer 9) G-3¢). Thebiotinylated RNA was immobilized on a streptavidin-coatedBiacore chip (SensorChip SA, BR-1003-98; General ElectricCompany) by an injection in binding buffer at a concentrationof 25 mg/mL (20 mM HEPES, pH 7.4, 150 mM NaCl, and 2 mMCaCl2) at 10mL/min. The RNA was refolded by heating to65�C followed by cooling to 37�C before immobilization. Tomeasure binding kinetics, 5 concentration of purified protein(prepared by serial dilutions from 250 to 15.6 nM) were in-jected at a flow rate of 15mL per minute. After binding, thesurface was regenerated by injecting 50 mM NaOH at a flowrate of 15mL per minute for 20 seconds. The KD values werecalculated by global fitting of the 6 concentrations of PSMAover a constant density of A9g aptamer (1001,1 RU). Thebinding data were fit to a 1:1 binding with a mass transfermodel to calculate kinetic parameters as previously described(Hernandez et al., 2009; Soontornworajit et al., 2011).

RNA structural modeling and PSMA docking

RNA 2-dimensional structures predictions. At the2-dimensional (2D) structural level, an RNA structure is de-scribed by the base pairs contained in the structure. The 2Dstructure of an RNA is predicted from the partition function,Q, defined as the sum over all the possible conformations:

Q¼ +s

e�DGS=kBT , where DGs is the free energy of a given

structure, s. The conformational sum +s

includes all the possible

secondary and pseudoknotted structures. The free energy foreach given structure, DGstacks, is determined from DGs =DGstacks - TDSloop, where DGstacks is the total free energy of thebase stacks as determined from the Turner rules (Serra andTurner, 1995), and - TDSloop is the loop free energy for thesecondary and pseudoknotted structures as determined fromthe Vfold model (Cao and Chen, 2005, 2006, 2009; Chen, 2008;Cao et al., 2010). To predict the 2D structures, the probability Pijof finding nucleotides i and j to form a base pair is computed. Pijis calculated from the conditional partition function Qij :Pij =Qij/Q. Here, Qij is the sum over all the possible conformationscontaining the (i, j) base pair. From the base pairing probabilitiesPij for all the possible (i, j) pairs, we predict the 2D structures.

302 ROCKEY ET AL.

G

G

G

AG

GA

C

G

A

U

G

CG

G

A

CC

GA

AAA A

G

A

CC

UG

A

C

UU

C UA

U

A

C

UAA

GU

C

U

A

C

GU

U

C

C

C

AGA

C

G

A

CU

C

C

C1

10

20

30

40

50

60

66

A9a(66mer)

A9b(60mer)

G

GG

A

U

GCG

G

ACC

GA

AA

A A

G

A

C

CU

G

A

C

UU

C UA

U

A

C

UAA

G

U

C

U

A

C

GU

U

C

C

C

AG

AC

C

C

1

10

20

30

40

50

55

A9c(55mer)

G

G

GA

C

G

A

U

G

CG

G

A

CC

GA

AA

AA

G

A

C

CU

G

A

C

UU

C U A

U

A

C

UA

A

G

U

C

U

A

C

GU

U

C

C

C

AGA

C

G

AC

C

C1

10

20

30

40

50

61

A9d(61mer)

G

GG

CG

G

ACC

GA

AA

A AG

A

CC

UG

A

C

UU

C UA

U

A

C

UAA

G

U

C

U

A

C

GU

U

C

CC

AC

C1

10

20

30

40

49

A9e(49mer)

A9f(51mer)

A9g(43mer)

G

G

G

A

G

GA

C

G

A

U

GCG

G

A

C

C

GA

AAA

A

G

A

C

CU

G

A

C

U

UC U A

U

A

C

UA

A

G

U

C

U

A

CG

U

U

C

C

C

AGA

C

G

A

CU

C

G

C

C

C

GA

1

10

20

30

40

50

60

70

A9(70mer)

*G

G

GA

C

G

A

U

G

CG

G

A

CC

GA

AA

A A

G

A

C

CU

G

A

C

UU

C UA

U

A

C

UA

A

G

U

CU

A

C

GU

U

C

C

C

AGA

C

G

C

C

C1

10

20

30

40

50

60* *

**

* **

**

** *

*G

G

G

C

G

G

ACC

GA

AAA

A

G

A

CC

UG

A

C

UU

C U AU

A

C

UA

A

G

U

C

U

A

C

GU

UC

CCA

G

C

C

C1

10

20

30

40

51**

*

*

G

G

G

AC

C

GA

AA

AAG

A

C

CU

G

A

C

UU

C UA

U

A

C

UA

A

G

U

C

U

A

C

G

UU

C

C

C1

10

20

30

40

43*

0

20

40

60

80

100

120

-RNA scrambled A9 A9a A9b A9c A9d A9e A9f A9g

NA

AL

AD

Ase

Act

ivit

y (%

)N

AA

LA

DA

se A

ctiv

ity

(%)

+ H2O NAALADaseNAAG NAA Glu+

0

20

40

60

80

100

120

140

-RNA A9 A10 A10-3 A10 scrambled A10-3.2 A10-3.2scrambled

NH N

H

OO

CO2-

CO2-

CO2-

CO2-

CO2-

CO2-

CO2-H2N

NH

+H2O

NAALADaseO

+

[3H]-NAAG NAA [3H]-Glu

A

B

C

FIG. 1. Functional characterization of various truncations of the A9 PSMA RNA aptamers generated by using RNAsecondary structural prediction algorithms. (A) RNA aptamers A9, A10, A10-3, and A10-3.2 were incubated with recom-binant PSMA protein. Production of [3H]-glutamate from [3H]-NAAG was measured by using an NAALADase assay. RNAaptamers A10 scrambled and A10-3.2-scrambled were used as negative controls in this assay. NAALADase activity in thepresence of each RNA was normalized to the no-RNA sample ( -RNA). (B) Secondary structural predictions of truncated A9aptamers generated using the RNAStructure 4.6 algorithm. Base changes are denoted by an asterisk (*). Base changes wereintroduced to retain a leading GGG transcription start codon at the 5¢ end of the truncated RNA sequences or to maintainbase complementarity at the 3¢ end. (C) Effect of A9 aptamer and truncated derivatives of the A9 aptamer on PSMAenzymatic activity. NAALADase activity was normalized as in part (A) above. (D) A9 and A9g RNA aptamers inhibit PSMANAALADase enzymatic activity with approximate IC50 values of 10 nM. PSMA, prostate-specific membrane antigen.

(Figure continued/)

303

RNA 3D structures predictions. The 3D structures of theRNAs were generated from the predicted 2D structures (Caoand Chen, 2011). The helices and loop/junctions in thestructure are identified from the 2D structures. For example,the A9g structure contains 2 helices P1 and P2 and an internalloop L1, a bulge loop C16, and a hairpin loop L2. P1 is the helixfrom base pair G1-C43 to base pair G7-C37, and P2 is the helixfrom base pair A12-U35 to base pair C15-G32. The internalloop L1 includes nucleotides from A8 to A11 and nucleotideA36. The hairpin loop includes nucleotides from G18 to A30.The 3D coordinates of the helices P1 and P2 were configuredby using A-form RNA helix coordinates. For the internal loop,bulge loop, and hairpin loop, the fragment-based method tosearch for the optimal template structures from the knownstructures in the PDB database was employed (Cao and Chen,2011). An optimal template is defined as the template withthe minimum substitution between the original loop and thetemplate sequence. For instance, the optimal template for theinternal loop L1 (5¢G7AAAA3¢, 5¢A36C3¢) was found to bethe loop (5¢AAAAA3¢, 5¢UA3¢) in the PDB structure 1J5A. Toachieve the optimal fit of the template structure, the terminalmismatch A11-A36 was placed within the helix P2. A 3Dscaffold structure was generated based on the helices and theloop template structures. In the last step, the 3D scaffoldstructure was further refined by using AMBER energy mini-mization (Case et al., 2005).

Predicting the RNA binding modes on PSMA. Thebinding modes of the RNA on the PSMA were constructed byusing our protein-RNA docking program. Specifically, thecrystal structure of PSMA was downloaded from the ProteinData Bank (PDB code: 1Z8L) (Davis et al., 2005). Water, ions,and ligands were removed from the protein. The modeledRNA 3D structure was used for the RNA. Then, the putativebinding modes of the RNA on PSMA were globally searchedby using our Fast Fourier Transform-based macromolecular

docking program MDockPP (Huang and Zou, 2010).MDockPP uses a hierarchical approach to construct the com-plexes between biological macromolecules. First, the proteinwas represented by a reduced model, in which each side chainon the protein surface was simplified and replaced by its centerof mass. Compared with the all-atom model, the reducedmodel allows larger side-chain flexibility during binding modesampling. Shape complementarity was used as a filtering cri-terion to generate several thousands of putative bindingmodes. These modes were further refined by our iterativelyderived knowledge-based scoring function ITScorePP(Huang and Zou, 2008) using the all-atom model to accountfor the atomic details. The top-ranked binding mode thatdoes not interfere with the putative membrane position andthe PSMA dimericinterface was selected as the predictedPSMA-RNA complex.

Cell culture

The PSMA-positive prostate cancer cell line 22Rv1(1.7) wasmaintained as described in Dassie et al. (2009) in RPMI 1640media with 10% FBS and 1% nonessential amino acids. ThePSMA-negative prostate cancer cell line (PC3) was main-tained according to the supplier’s recommendations (ATCC#CRL-1435) in DMEM/F12 media with 10% FBS. Cells weremaintained at 37�C with an atmosphere containing 5% CO2.

Cell binding assay

One day before the binding assay, cells were plated in a 24-well plate at a density of *100,000 cells per well. All subse-quent procedures were performed on ice to prevent aptamerinternalization. Before binding, each well was washed twicewith 1 mL of ice-cold Dulbecco’s phosphate-buffered salinein the absence of divalent cations (DPBS -/-) to removegrowth media. Aptamers were 5¢ end-labeled with 32P usingPNK from New England Biolabs as previously described(McNamara et al., 2008). The concentration of 32P-radiolabeledaptamer was measured with UV-visible absorption spectros-copy, and serial dilutions ranging from 1000 to 0 nM wereperformed. To measure nonspecific binding, serial dilutionswere also made containing a high fixed concentration ofnonradiolabeled A9g aptamer, at 10mM (10,000 nM). Both setsof dilutions were incubated with the cells in the 24-well plateon ice in a volume of 100mL. After 1 hour, the binding reactionmixture was aspirated off the cells, and the cells were washedtwice with 0.5 mL of ice-cold DPBS. Bound RNA was collectedby washing with 0.5 mL of 0.5 N NaOH that was added to3 mL of scintillation fluid, and activity was measured. Foreach dilution, specific binding was calculated by subtractingthe activity of the sample with a high concentration of non-radiolabeled (‘‘cold’’) aptamer added (ie, nonspecific binding)from the sample without cold aptamer added (ie, total bind-ing). The data were plotted and fit to a one-site saturationbinding model by using the nonlinear regression algorithmof the software package Sigma Plot. Experiments were per-formed in duplicate.

Cell internalization assays

22Rv1(1.7) PSMA-positive prostate cancer cells (target) andPC-3 PSMA-negative prostate cancer cells (nontarget) weregrown to confluency in a 6-well plate. Cells were washed

log [RNA pM]

0 2 4 6

No

rmal

ized

CP

M

0.0

0.2

0.4

0.6

0.8

1.0

1.2

A9A9g

D

FIG. 1. (Continued).

304 ROCKEY ET AL.

twice with 1 mL of DPBS prewarmed at 37�C. Cells were thenblocked with 1 mL of 100mg/mL yeast tRNA prewarmed at37�C. After 15 minutes, the block was removed, and 100 pmolRNA aptamer in DPBS was added to cells for 30 minutes at37�C with 5% CO2. Cells were washed once with ice-coldDPBS followed by 2 washes of ice-cold 0.5M NaCl in DPBS.The internalized RNA was recovered by using TRIzol reagent.Quantitative reverse transcription (RT)-PCR was performedby using the iScript One-Step RT-PCR Kit with SYBR Green(Cat# 170-8893) from Bio-Rad Laboratories. Samples werenormalized to an internal RNA reference control. Specifically,0.5 pmol/sample m12-23 aptamer (McNamara et al., 2008)was added to each sample along with TRIzol as a referencecontrol. Primer sets included the internal reference primer setfor m12-23 (Sel1), the A9g primer set (amplifies A9, A9g, andA9g.6), the A10 primer set (amplifies A10 and A10-3.2), andthe A10-3.2 scrambled primer set. Samples were first nor-malized to the internal reference RNA (m12-23) and then ac-cording to the relative amount of RNA internalized vs. thenontarget control cells (PC3).

Primer sequences for the quantitative RT-PCR are as fol-lows: Sel1 5¢ primer: 5¢-GGGGGAATTCTAATACGACTCACTATAGG GAGAGAGGAAGAGGGATGGG-3¢; Sel1 3¢primer 5¢-GGGGGGATCCAGTACTATCGACCTCT GGGTTATG-3¢; A9g 5¢ primer: 5¢-TAATACGACTCACTATAGGGACCGAAAAAGACC-3¢; A9g 3¢ primer:5¢-GGGAACGTAGACTTAG-3¢; A10 5¢ primer: 5¢-TAATACGACTCACTATAGGGAGGA CGATGCGG-3¢; A10-3.2 3¢ primer: 5¢-AGGAGTGACGTAAACATG -3¢; A10-3.2 scrambled 5¢ primer: 5¢-TAATACGACTCACTATAGGGGCATGCCTAGCT-3¢; A10-3.2scrambled 3¢ primer: 5¢-CCGCGCATAAGCCATGGG-3¢.

Results

Rational truncation of A9 PSMA RNA aptamer

The PSMA RNA aptamers A9 and A10 have been selectedfor their ability to inhibit PSMA’s enzymatic activity (Lupoldet al., 2002). Since PSMA’s enzymatic activity has been im-plicated in carcinogenesis (metastatic potential) (Lapiduset al., 2000), optimized, truncated versions of these inhibi-tors promise to be valuable agents not only for targeted im-aging and therapy of prostate cancer but also to directlyinhibit PSMA’s pro-metastatic functions. We used the NAA-LADase assay to assess the inhibitory activity of previouslydescribed, truncated versions of the A10 RNA aptamer:A10-3 (56 mer) (Lupold et al., 2002) and A10-3.2 (39 mer)(Fig. 1A). The NAALADase activity of PSMA hydrolyzesN-acetylaspartylglutamate (NAAG) to N-acetylaspartate andglutamate (Fig. 1A; insert). As previously described (Lupoldet al., 2002), A10-3 retains NAALADase inhibitory activity,albeit less efficiently compared with the full-length A10and A9 RNA aptamers. In contrast, A10-3.2 (39 mer) hadno NAALADase inhibitory activity. This was confirmed athigher RNA concentrations up to 3.8 mM (data not shown).Scrambled versions of the A10 and A10-3.2 aptamers wereused as negative controls in this assay. These scrambledaptamers have the same number of nucleotides and basecomposition as their wild-type counterparts but possess a‘‘scrambled’’ sequence.

As previously described, A9 is a better inhibitor of PSMAenzymatic activity compared with A10 (Lupold et al., 2002).Thus, we set out to determine the NAALADase inhibitory

activity of various truncations of the A9 aptamer. Previousattempts at truncating the A9 aptamer have proved unsuc-cessful (Lupold et al., 2002). Thus, rather than performing aseries of base deletions from the 3¢ end, we reasoned thatmaintaining the overall structure of the PSMA-interactingregion of the aptamer would be essential for retaining activity.To this end, a series of 5¢ and 3¢-end base deletions were made,and the RNA secondary-structure prediction programRNAStructure 4.6 was used to select those truncations thatretained the predicted secondary structural motifs of the full-length A9 aptamer (Fig. 1B). In addition, selective basechanges were made at the 5¢ and 3¢ ends to maintain aT7transcription start-site (5¢GGG) and maintain base-paringcomplementarity at the 3¢ end.

Seven initial truncated versions of the A9 aptamer weredesigned (A9a through A9g) with lengths ranging from 66bases (A9a) to 43 bases (A9g). The NAALADase assay wasused to assess inhibition of PSMA enzymatic activity by thevarious truncations. A scrambled RNA aptamer sequence(71 mer) did not inhibit enzymatic activity. Remarkably, all7 truncations inhibited PSMA NAALADase activity as wellas full-length A9 under these assay conditions (800 nMRNA) (Fig. 1C). We next determined the inhibitory potencyof the shortest truncation, A9g (43 mer) compared with thefull-length A9aptamer. Inhibition was tested over a range ofRNA concentrations (20 pM to 800 nM). Both A9g (43 mer)and A9 (70 mer) inhibited NAALADase activity with anIC50 of 10 nM under the assay conditions (Fig. 1D), thussuggesting that A9g, similar to A9, retains key structural/sequence elements important for inhibition of PSMA enzy-matic activity.

A second series of truncations were made in an attempt tofurther decrease the length of the A9g aptamer and to assessstructural and sequence elements important for PSMA inhi-bition (Fig. 2A). The truncations A9h (37 mer) and A9i (24mer) retain sequence and structural loop elements of A9g,whereas A9j (30 mer) and A9k (21 mer) retain sequence andstructural stem elements of A9g (Fig. 2A). Interestingly, unlikeA9 and A9g, none of these additional truncations (A9h-A9k)exhibited inhibitory activity under the assay conditions(200 nM RNA concentration) (Fig. 2B). Together, these resultssuggest that key sequence and/or structural elements forPSMA inhibition are present within bases 1–43 of the A9gaptamer.

A9g binds to PSMA with high affinity and specificity

The NAALADase activity assay provides an indirectmeasurement of the interaction of the PSMA aptamers withPSMA. To determine the binding profile of the A9g aptamerfor PSMA, we performed filter-binding assays (Fig. 3A) and(SPR/BIACore) with recombinant, purified human PSMAprotein (Fig. 3B). As determined by the filter binding assay,the A9g aptamer retains the same binding profile as the full-length A9 (Fig. 3A). A more extensive measure of binding byanalyzing kinetic interaction data using SPR/BIACore wasalso performed. In these experiments, biotinylated A9g RNAwas immobilized on streptavidin-coated gold chips. A solu-tion containing the analyte of interest (recombinant purifiedPSMA protein) was injected over the chip during an associa-tion phase, thus allowing for measurement of the binding on-rate (kon). After the injection was halted, the rate of

COMPUTER-GUIDED OPTIMIZATION OF AN RNA APTAMER 305

dissociation (koff) was measured. By repeating these mea-surements at various analyte (PSMA) concentrations, an ac-curate estimation of binding was determined (KD = koff/kon).The KD of A9g for PSMA ranged from 5 nM to 30 nM intriplicate experiments (lowest value shown) (Fig. 3B). Thedifferences in the absolute KD values obtained by filter bind-ing and SPR are likely due to the intrinsic differences withregard to these assays (Arraiano et al., 2008).

Structure-function analysis of A9g binding to PSMA

A series of base changes were introduced within A9g in anattempt to identify the sequence/structural elements neces-sary for binding to PSMA. Inherent in these experiments is theassumption that the base changes only create local changes inthe RNA structure and not a global change in folding. Forthese experiments, the A9g aptamer was divided into 2 stemregions (S1 and S2) and 3 loop regions (L1, L2 and L3) (Fig. 4A).Base changes were made to either preserve or disrupt thesevarious structural elements. The RNA-secondary structureprediction algorithm, RNAStructure 4.6, was used to predictfolding of the modified A9g RNAs (A9g.1-A9g.6).

To address the importance of the S2 stem sequence, the A-Ubase pair in the stem region S2 was replaced with either a G-Cor a C-G base pair (A9g.1 and A9g.2 respectively) (Fig. 4A).A9g.1 and A9g.2 were predicted to retain the overall sec-

ondary structure as A9g (Fig. 4A). As predicted, A9g.1 andA9g.2 resulted in RNA aptamers with comparable inhibitoryactivity as A9g (Fig. 4B). In contrast, a base change within S2that was predicted to lengthen the stem (A9g.5) resulted in aloss of PSMA inhibitory activity, thus suggesting that theoverall structural and not sequence elements of S2 are im-portant for the RNA’s inhibitory function. We next addressedthe importance of each loop (L1, L2, and L3) by introducingbase changes that would disrupt the predicted folding of theloops (A9g.3, A9g.4, and A9g.6 respectively). With the ex-ception of A9g.4, all base changes completely abrogated theability of the RNA aptamers to inhibit PSMA enzymatic ac-tivity (Fig. 4B), thus suggesting that the loops are required forfunction. In the case of A9g.4, inhibitory activity was de-creased by*50% compared with A9g. Interestingly, 2 distinctsecondary structures (A9g.4a and A9g.4b) with similar mini-mum free energies (DGs) were predicted for A9g.4 (Fig. 4A).The predicted free energies of these 2 structures were - 9.9and - 9.4 kcal$mol - 1, respectively. To assess whether loss ofinhibitory function correlates with loss of binding to PSMA,we performed filter binding assays to determine binding ofA9g.3-A9g.6 to recombinant PSMA (Fig. 4C). With the ex-ception of A9g.4, the binding capacity (Bmax) of PSMA forthese mutants was severely diminished. The binding of A9g.4mirrored its inhibitory activity (Fig. 4B), with a binding ca-pacity for PSMA of *50% compared with A9g.

A9h A9i A9j A9k

GGGC

CU

G

A

C

UU

C U AU

A

C

UA

A

GCCC1

10

20

24*

* *

*

GGGA

CCG

AAA

A AGAC

CU

A

GU

CU

A

CGU

UCCC1

10

20

30* GGGA

CCG

A

AA A A

U

ACGU

UCCC1

10

2021*

0

20

40

60

80

100

120

-RNA scrambled A9 A9g A9h A9i A9j A9k

NA

AL

AD

ase

Act

ivit

y (%

)

A9g

G

G

G

AC

C

GA

AA

AAG

A

C

CU

G

A

C

UU

C UA

U

A

C

UA

A

G

U

C

U

A

C

G

UU

C

C

C1

10

20

30

40

43*

+ H2O NAALADaseNAAG NAA Glu+

G

G

GG

AAA

AA

G

A

CC

UG

A

C

UU C U

A

U

A

C

UAA

G

U

C

U

A

CC

C

C1

10

20

30

37*** *

**

B

A

FIG. 2. Further truncation of the A9 aptamer causes loss of inhibitory activity. (A) Secondary structural predictions oftruncated A9 aptamers generated using the RNAStructure 4.6 algorithm. Base changes are denoted by an asterisk (*). Basechanges were introduced to retain a leading GGG transcription start codon at the 5¢ end of the truncated RNA sequences or tomaintain base complementarity at the 3¢ end. (B) Effect of A9 aptamer and truncated derivatives of the A9 aptamer on PSMAenzymatic activity. NAALADase activity was normalized as in Fig. 1.

306 ROCKEY ET AL.

Assessment of binding specificity of A9g to PSMA

Binding specificity of the A9g aptamer for PSMA was de-termined by using SPR/BIACore (Fig. 4D; left panel). Bindingspecificity was assessed by comparing the kon and koff rates ofA9g for recombinant PSMA protein (target) to the Kon and koffrates of A9g for nontarget proteins (BSA and HER2). For theseexperiments, biotinylated A9g RNA was immobilized onstreptavidin-coated gold chips. No appreciable interactionbetween A9g and the nontarget proteins (BSA and HER2) wasmeasured (Fig. 4D; left panel). Lack of binding of A9g.6 toPSMA was also confirmed with SPR/BIACore (Fig. 4D; rightpanel). In addition, there was no measurable binding of A9g.6to the nontarget proteins (BSA and HER2). These data provideconfirmation of binding specificity of A9g for PSMA (Fig. 4D).

RNA tertiary structure predictions and RNA-proteindocking studies

With the exceptions of A9g.1 and A9g.2 that were designedto have the same secondary structure as wild-type A9g, all theother A9g-derivatives experienced a significant decrease in theirability to inhibit and bind PSMA. It may be that each of thepredicted secondary structural elements examined play a rolein the aptamer’s binding to PSMA. Alternatively, any of thechanges made to the predicted structural elements may disruptthe ‘‘global’’ folding of the RNA, thus rendering it inactive.

To provide additional insight into the interaction of the A9gRNA aptamer with PSMA, a tertiary structure model of A9gwas created. The predicted tertiary structure of A9g wascomputationally docked to a crystal structure of PSMA (Daviset al., 2005) (Fig. 5A; left panel). Interestingly, the RNA-pro-tein docking analysis revealed 2 bases, adenosine at position 9(A9) and uridine at position 39 (U39), that were predicted tointeract directly with PSMA. The amine group of A9 forms ahydrogen bond with a backbone carbonyl of PSMA, and U39forms multiple close van der Waals interactions with PSMAside chains. On the basis of these predictions, base changeswere made to retain the hydrogen bond at position A9 (Fig.5A; compare middle and right panels) and to test the necessityof U at position 39. Specifically, the uridine at position 39 wasreplaced with either an adenosine (A9g.7; U39A) or a cytosine(A9g.8; U39C), and the adenosine at position 9 was replacedwith a cytosine (A9g.9; A9C) (Fig. 5A; right panel). Predictedsecondary structures for these A9g variants are shown in Fig.5B. Not surprisingly, the A9g (A9C) variant retained PSMAinhibitory activity, albeit less effectively compared with A9g(Fig. 5B). In contrast, the A9g (U39A), A9g (U39C), and A9g(U39G) variants completely lost inhibitory activity (Fig. 5B).Notably, unlike the A9g (U39G) variant (identical to A9g.3,Fig. 4A), theA9g (U39A) and A9g (U39C) variants were notpredicted to alter the secondary structure of A9g (Fig. 5B).These data suggest that sequence conservation (uridine) atposition 39 may be more important than the overall structureof the L1 loop for conferring the RNA aptamer’s inhibitoryfunction.

Based on the data just provided, we hypothesized that afurther truncation of A9g which retains suridine at position 39should result in an RNA aptamer with comparable PSMAinhibitory activity to A9g. To test this hypothesis, we removedthe most distal G-C base-pair of A9g (A9L; 41 mer). We alsointroduced a base change at the first position to maintain the5¢-GGG T7 RNA polymerase transcription start (Fig. 5C; leftpanel). As predicted, A9L was equally as effective as A9g atinhibiting PSMA enzymatic activity (Fig. 5C; right panel).Elimination of additional bases from the 5¢ or 3¢ termini (eg,A9h; 37 mer) abrogated inhibition of PSMA enzymatic ac-tivity (Fig. 5C; right panel). These findings were consistentwith altered folding of these shorter RNAs as predicted byusing the RNA secondary structure prediction algorithm(RNAStructure 4.6) and loss of sequence elements (eg, U atposition 39) required for function.

A9g and A9L bind to and internalize into PSMA-positiveprostate cancer cells

Binding of A9g to PSMA expressed on the surface ofprostate cancer cells was confirmed by incubating varyingamounts of 32P-labeled A9g with either PSMA-positive

FIG. 3. Binding of A9 and A9g to human PSMA. (A) Asaturation filter binding assay was used to measure bindingof A9 and A9g to recombinant human PSMA protein. Thecalculated KD for A9 was 110 nM, and the KD for A9g was130 nM. The fraction bound was normalized to the Bmax(maximal binding capacity) of A9. (B) Measurement of thebinding affinity of A9g for recombinant human PSMA pro-tein by surface plasmon resonance (SPR, BIACore). The datawere fit to a 1:1 binding with a mass transfer model. The KDof A9g calculated from the model was 5 nM with an w2 valueof 1.51. The on-rate (ka) was 1.15 · 104 M - 1$s - 1, and the off-rate (kd) was 5.7 · 10 - 5 seconds - 1.

COMPUTER-GUIDED OPTIMIZATION OF AN RNA APTAMER 307

http://www.liebertonline.com/action/showImage?doi=10.1089/nat.2011.0313&iName=master.img-000.jpg&w=237&h=366

(22Rv1 clone 1.7) (Dassie et al., 2009) or PSMA-negative(PC-3) prostate cancer cells on ice (to prevent internaliza-tion into the cells) (Supplementary Fig. S1; SupplementaryData are available online at www.liebertonline.com/nat).The PSMA-expressing cells were found to have a higher

binding capacity for A9g compared with the PSMA-negative cells (Supplementary Fig. S1). The backgroundbinding to PC-3 cells is thought to be a result of free 32Pafter exo-nuclease digestion on the cell surface (datanot shown).

FIG. 4. Characterization of A9g binding to PSMA. (A) Secondary structural predictions of truncated A9 aptamers generatedusing the RNAStructure 4.6 algorithm. Base changes are denoted by an asterisk (*). Base changes were introduced in an attemptto either retain the predicted secondary structure (A9g.1 and A9g.2) or disrupt various secondary structural elements (A9g.3-A9g.6) of A9g. Two secondary structure predictions were given for the A9g.4 sequence, denoted by A9g.4a and A9g.4b. (B)Effect of A9g aptamer derivatives (A9g.1 through A9g.6) on PSMA NAALADase inhibitor activity. NAALADase activity wasmeasured and normalized as in Fig. 1. (C) Saturation filter binding assay of A9g aptamer and A9g aptamer derivatives (A9g.3-A9g.6). (D) Binding of A9g to recombinant human PSMA, recombinant rat HER2 (rHER2), and BSA using BIACore (left panel).(E) Binding of A9g.6 to recombinant human PSMA, recombinant rat HER2 (rHER2), and BSA using BIACore (right panel).

308 ROCKEY ET AL.

http://www.liebertonline.com/action/showImage?doi=10.1089/nat.2011.0313&iName=master.img-001.jpg&w=488&h=518

Aptamers that bind to cell-surface proteins (eg, cancerepitopes) can be developed for imaging applications (Hickeet al., 2006). In addition, aptamers with cell-internalizingproperties can be harnessed for delivery of therapeutic agentsinto target cells (Bagalkot et al., 2006; Dhar et al., 2008; Gu

et al., 2008; Cao et al., 2009). Both the A9 and A10 RNA ap-tamers were demonstrated to be effective at delivering cargosthat require internalization, such as cytotoxic drugs (Far-okhzad et al., 2004) and siRNAs (Chu et al., 2006; McNamaraet al., 2006). For therapeutic development, the A10 aptamer

FIG. 5. Truncated A9 PSMA aptamers derived based on RNA tertiary structure and protein/RNA docking predictions. (A)Modeled tertiary structure of A9g docked to a crystal structure of PSMA. The bases A9 and U39 are predicted to form directinteractions with the crystal structure of PSMA. The amine group of A9 is predicted to form a hydrogen bond with abackbone carbonyl of PSMA (close up; middle panel). Right panel; close up of A9g (A9C) variant where the A at position 9was changed to a C to retain the hydrogen bond. (B) Secondary structural predictions of A9g aptamer and A9g aptamerderivatives generated using the RNAStructure 4.6 algorithm (left panel). Base changes are denoted by an asterisk (*). Secondarystructural predictions of A9g were generated to test the importance of the uracil at position 39 and the adenosine at position 9.Effect of A9g and A9g aptamer derivatives (U39A, U39C, U39G, and A9C) on PSMA NAALADase activity (right panel). (C)Secondary structural predictions of A9g aptamer and truncated derivatives A9L (41 mer) and A9h (37 mer) using theRNAStructure 4.6 algorithm (left panel). Base changes are denoted by an asterisk (*). Effect of A9L (41 mer) and A9h (37 mer)aptamers on PSMA NAALADase activity. NAALADase activity was measured and normalized as in Fig. 1 (right panel).

COMPUTER-GUIDED OPTIMIZATION OF AN RNA APTAMER 309

http://www.liebertonline.com/action/showImage?doi=10.1089/nat.2011.0313&iName=master.img-002.jpg&w=486&h=482

was further truncated to 39 bases (A10-3.2) while retaining theability to bind to PSMA on the surface of cells and deliver itstherapeutic siRNA cargo into PSMA-expressing prostatecancer cells (Dassie et al., 2009). Unfortunately, the shorterA10-3.2 aptamer no longer exhibits PSMA inhibitory activity(Fig. 1A). Since inhibitory activity, binding, and internaliza-tion ability do not necessarily coincide, we performed an in-ternalization assay to assess whether the shorter A9 aptamervariants (A9g and A9L), which retain PSMA inhibitory ac-tivity (Fig. 5C), internalize into PSMA-expressing prostatecancer cells (Fig. 6A). Full-length A9, A9g (43 mer), and A9L(41mer) aptamers were incubated with either PSMA-positive(22Rv1 clone 1.7) or PSMA-negative (PC-3) prostate cancercells at 37�C to enable cell internalization. Cells were washedwith a high-salt wash buffer containing 0.5 M NaCl to removenonbinders or aptamers bound to the surface of the cells. In-ternalized aptamers were recovered by Trizol extraction. Theefficiency of internalization for each RNA aptamer was as-sessed by using quantitative RT-PCR (Fig. 6). No loss in in-ternalization ability was observed for the truncated A9variants (A9g and A9L) compared with the full-length A9RNA aptamer (Fig. 6A). As expected, A9g and A9L retainedspecificity for cells expressing PSMA (Fig. 6A). Importantly,A9g and A9L internalized more efficiently into PSMA ex-pressing prostate cancer cells compared with A10 and the A10truncated variants (A10-3 and A10-3.2) (Fig. 6B). No inter-nalization was observed with a scrambled A10-3.2 aptamersequence or with a functionally inactive mutant of A9g(A9g.6) (Fig. 6). All A10 and A9 RNA aptamer derivativesretained specificity for PSMA expressing cells (22Rv1 clone1.7) compared with PSMA-negative cells (PC-3) (Fig. 6B). Thefold increase of RNA recovered from PSMA-expressing cellsversus RNA recovered from PC-3 cells is shown for each RNAaptamer. No statistically significant difference in internaliza-

tion is observed for A10 and A10-3.2 (P = 0.1). In contrast, thetruncated A9 variants (A9g and A9L) internalized more effi-ciently into PSMA-expressing cells compared with either thefull-length A9 aptamer (P < 0.1) or A10 aptamers. This couldbe a result of steric hindrance or interaction of a part of theaptamer with other cellular factors that may hinder or retarduptake (data not shown). Together, these data confirm thatthe truncated A9 aptamer variants (A9g and A9L) retain tar-get-specific cell internalizing properties and can, thus, be de-veloped into effective targeted delivery agents for prostatecancer.

Discussion

Here, we describe a ‘‘rational truncation’’ approach thattakes advantage of computer-generated RNA structuremodels to facilitate the truncation of RNA aptamer sequencespostselection. This approach enabled us to engineer truncatedversions of the PSMA A9 aptamer (Lupold et al., 2002) thatretain binding affinity, specificity, and functionality. Com-puter-generated RNA secondary structure models were usedto remove bases from both the 5¢- and 3¢- termini of the RNAand introduce base changes to conserve those secondarystructural elements that are predicted to be necessary forbinding to PSMA. This analysis resulted in a 27-base trunca-tion of the PSMA A9 RNA aptamer, yielding an RNA oligo-nucleotide of 43 nucleotides in length (A9g), which binds torecombinant PSMA with nanomolar affinity (KD = 5 nM) (Fig.3B) and retains PSMA inhibitory activity (Fig. 1D). Im-portantly, we show that similar to A9, A9g retains the abilityto internalize into PSMA-expressing prostate cancer cells (Fig.6) and, thus, could be used for targeted delivery of therapeuticagents (toxins, siRNAs, and radionuclides). In addition tocomputer-generated RNA secondary structure models, we

0

0.5

1

1.5

2

2.5

3

3.5

A9 A9g A9L A9g.6

No

rmal

ized

RN

A R

eco

very

0

0.5

1

1.5

2

2.5

3

3.5

A9 A9g A9L A9g.6

No

rmal

ized

RN

A R

eco

very

22Rv1(1.7)

PC3

0

0.5

1

1.5

2

2.5

3

3.5

A9 A9g A9L A9g.6 A10 A10-3.2

A10-3.2 scr

No

rmal

ized

RN

A R

eco

very

130 80

5

36

40

0

0.5

1

1.5

2

2.5

3

3.5

- -3.2

No

rmal

ized

RN

A R

eco

very

86

14

22Rv1(1.7)

PC3

A B

FIG. 6. Truncated A9 aptamers bind to and internalize into PSMA expressing cells. (A) Internalization of PSMA RNAaptamers A9, A9g (43 mer), A9L (41 mer), and A9g.6 into prostate cancer cells expressing PSMA. Internalization wasmeasured by using quantitative reverse transcription-polymerase chain reaction. RNA recovery was normalized to recoveryof an internal RNA control. (B) Internalization of PSMA RNA aptamers A10, A9, and derivatives into PSMA expressingprostate cancer cells. A10-3.2 scrambled and A9g.6 aptamers were used as negative controls for internalization in this assay.The fold enrichment in recovery with regard to non-PSMA expressing cells is reported.

310 ROCKEY ET AL.

combined predictive RNA tertiary structure models withprotein docking studies to obtain further insights into theA9g-PSMA interaction (Fig. 5). This analysis revealed keynucleotides within A9g critical for binding to PSMA (Fig. 5A).Further, this analysis enabled us to perform an additional 2-nucleotide truncation of A9g, thus resulting in a 41-nucleo-tide-long RNA oligonucleotide (A9L) with comparable bind-ing affinity and activity to A9 and A9g (Fig. 5C).

The successful truncation of the A9 PSMA aptamer is ofimportance in the light of recent data directly implicatingPSMA’s enzymatic activity in promoting carcinogenesis(Lapidus et al., 2000; Yao et al., 2010). PSMA has multiplecatalytic activities, including NAALADase, folatecarbox-ypeptidase, and dipeptidyl peptidase IV activity (Bacich et al.,2001). Recent studies have suggested a role for PSMA enzy-matic activity in cell migration and activation of oncogenicpathways (Lapidus et al., 2000; Yao et al., 2010). Importantly,inhibition of PSMA enzymatic activity by small molecule in-hibitors abrogates PSMA-mediated carcinogenesis ((Kularatneet al., 2009; Yao et al., 2010) and our unpublished data). Here,we show that the A9g (43 mer) and A9L (41 mer) aptamers,similar to A9, retain the ability to inhibit PSMA’s NAALA-Dase activity (Fig. 5C) and, thus, could be employed as thera-peutic inhibitors of PSMA. In contrast, a previously describedtruncated version of the A10 PSMA aptamer (A10-3.2; 39 mer),which retains binding to PSMA (Dassie et al., 2009), is unable toinhibit PSMA NAALADase activity (Fig. 1A).

The A10-3.2 aptamer has been successfully used by us todeliver siRNAs targeting cancer prosurvival genes to PSMA-expressing prostate cancer cells (Dassie et al., 2009). In thiscontext, the truncated aptamer serves solely as a delivery toolfor the therapeutic siRNA cargo. In principle, conjugation oftherapeutic siRNAs to the A9g and A9L aptamers, which wedemonstrate internalize efficiently and specifically intoPSMA-expressing cells (Fig. 6), could result in dual function-targeted reagents that are capable of inhibiting multiplecarcinogenic pathways (PSMA and prosurvivalgenes). Anaptamer-siRNA conjugate with dual function has been pre-viously described for the treatment of HIV infected cells(Zhou et al., 2008). In this article, an inhibitory aptameragainst gp120 was tethered to an siRNA against tat/rev, 2 viralgenes that drive replication of the virus. The aptamer-siRNAcombination reduced HIV infectivity and replication in cul-tured T cells (Zhou et al., 2008) and suppressed HIV-1 viralloads reversing CD4 + T cell decline in a humanized mousemodel of HIV (Neff et al., 2011).

In principle, the information provided by the theoreticalsecondary and tertiary RNA structure models can be used notonly to guide in the truncation of long RNA oligonucleotidesequences (as described herein) but also to enable the modifi-cation of key nucleotides to improve overall aptamer qualityand function (Zhou et al., 2011). Although large-scale, high-quality cGMP-grade (Current Good Manufacturing Process)synthesis of long RNA oligonucleotide aptamers (60–100 nu-cleotides long) remains a rate limiting step to their therapeuticpotential (Reese, 2005), other in vivo properties of these RNAs,such as their pharmacokinetics (PK) and pharmacodynamics(PD), can also hinder their therapeutic utility [reviewed in(Keefe et al., 2010)]. Several ways to optimize the PK/PD ofaptamers have been described. These include (1) the use ofmodified nucleotides that impart nuclease resistance, thus re-sulting in RNA aptamers with longer half lives in the blood

(Lin et al., 1994; Ruckman et al., 1998) and (2) chemical conju-gation of high-molecular-weight molecules (eg, 20–40 kDaPEG) to prevent exclusion by renal filtration (Kawaguchi et al.,1995; Watson et al., 2000; Healy et al., 2004). Although 2’-fluoromodified pyrimidines are usually incorporated into RNA ap-tamers during the selection process, additional modificationsare introduced postselection, using an atrial-and-error ap-proach that is laborious and is not guaranteed to work for allaptamers (Ruckman et al., 1998; Floege et al., 1999; Adler et al.,2008). In principle, theoretical RNA structure algorithms sim-ilar to the ones described herein can be utilized to identify basesthat when modified (with synthetic bases) may increase theoverall thermodynamic stability and nuclease resistance ofthese RNA aptamers without loss of function. Likewise, thesealgorithms can be used to identify critical residues that cannottolerate modifications (Fig. 5A).

In conclusion, our studies highlight the utility of theoreticalRNA secondary and tertiary structure models and proteindocking studies for guiding the truncation of RNA aptamersto enable and expedite large-scale chemical synthesis of theseRNAs for clinical applications. Importantly, these efforts haveresulted in a truncated PSMA A9 aptamer that due to itsshorter sequence length is now amenable to large-scale che-mical synthesis for targeted therapeutic applications in thesetting of prostate cancer. Finally, the ability to directly testthe computer-generated structural predictions by using ro-bust functional assays (binding and enzymatic activity) canenable the refinement of current RNA prediction algorithms.Once refined, these theoretical models can be applied to op-timize other aptamers with therapeutic potential.

Acknowledgments

The authors thank Dr. Luiza Hernandez for careful editingof this article. This work was supported by funding from theNational Institutes of Health [1RO1 CA138503-01 and1R21DE019953-01 to PHG; GM063732 to SJC; R21GM088517to XQ]; the National Science Foundation [MCB0920411,MCB0920067 to SJC; NSF CAREER Award DBI-0953839 toXZ]; the Roy J. Carver Charitable Trust [RJCCT 01-224 to PHG];and the RSNA Research Resident Grant [RR0905 to WMR].

Disclosure Statement

No competing financial interests exist.

References

ADLER, A., FORSTER, N., HOMANN, M., and GORINGER,H.U. (2008). Post-SELEX chemical optimization of a trypano-some-specific RNA aptamer. Comb. Chem. High ThroughputScreen. 11, 16–23.

ARRAIANO, C.M., BARBAS, A., and AMBLAR, M. (2008).Characterizing ribonucleases in vitro examples of synergiesbetween biochemical and structural analysis. Methods En-zymol. 447, 131–160.

BACICH, D.J., PINTO, J.T., TONG, W.P., and HESTON, W.D. (2001).Cloning, expression, genomic localization, and enzymatic activi-ties of the mouse homolog of prostate-specific membrane antigen/NAALADase/folate hydrolase. Mamm. Genome 12, 117–123.

BAGALKOT, V., FAROKHZAD, O.C., LANGER, R., and JON,S. (2006). An aptamer-doxorubicin physical conjugate as anovel targeted drug-delivery platform. Angew. Chem. Int. Ed.Engl. 45, 8149–8152.

COMPUTER-GUIDED OPTIMIZATION OF AN RNA APTAMER 311

BIESECKER, G., DIHEL, L., ENNEY, K., and BENDELE, R.A.(1999). Derivation of RNA aptamer inhibitors of humancomplement C5. Immunopharmacology 42, 219–230.

BUFF, M.C., SCHAFER, F., WULFFEN, B., MULLER, J.,POTZSCH, B., HECKEL, A., and MAYER, G. (2010). Depen-dence of aptamer activity on opposed terminal extensions:improvement of light-regulation efficiency. Nucleic Acids Res.38, 2111–2118.

BURGSTALLER, P., KOCHOYAN, M., and FAMULOK, M.(1995). Structural probing and damage selection of citrulline-and arginine-specific RNA aptamers identify base positionsrequired for binding. Nucleic Acids Res. 23, 4769–4776.

BURMEISTER, P.E., LEWIS, S.D., SILVA, R.F., PREISS, J.R.,HORWITZ, L.R., PENDERGRAST, P.S., MCCAULEY, T.G.,KURZ, J.C., EPSTEIN, D.M., WILSON, C., and KEEFE, A.D.(2005). Direct in vitro selection of a 2’-O-methyl aptamer toVEGF. Chem. Biol. 12, 25–33.

BURMEISTER, P.E., WANG, C., KILLOUGH, J.R., LEWIS, S.D.,HORWITZ, L.R., FERGUSON, A., THOMPSON, K.M., PEN-DERGRAST, P.S., MCCAULEY, T.G., KURZ, M., DIENER, J.,CLOAD, S.T., WILSON, C., and KEEFE, A.D. (2006). 2’-Deoxypurine, 2’-O-methyl pyrimidine (dRmY) aptamers as candi-date therapeutics. Oligonucleotides 16, 337–351.

CAO, S., and CHEN, S.J. (2005). Predicting RNA folding ther-modynamics with a reduced chain representation model.RNA 11, 1884–1897.

CAO, S., and CHEN, S.J. (2006). Free energy landscapes ofRNA/RNA complexes: with applications to snRNA com-plexes in spliceosomes. J. Mol. Biol. 357, 292–312.

CAO, S., and CHEN, S.J. (2009). A new computational approachfor mechanical folding kinetics of RNA hairpins. Biophys. J.96, 4024–4034.

CAO, S., and CHEN, S.J. (2011). Physics-based de novo pre-diction of RNA 3D structures. J. Phys. Chem. B 115, 4216–4226.

CAO, S., FURTIG, B., SCHWALBE, H., and CHEN, S.J. (2010).Folding kinetics for the conformational switch between alter-native RNA structures. J. Phys. Chem. B 114, 13609–13615.

CAO, Z., TONG, R., MISHRA, A., XU, W., WONG, G.C.,CHENG, J., and LU, Y. (2009). Reversible cell-specific drugdelivery with aptamer-functionalized liposomes. Angew.Chem. Int. Ed. Engl. 48, 6494–6498.

CASE, D.A., CHEATHAM, T.E., 3rd, DARDEN, T., GOHLKE,H., LUO, R., MERZ, K.M., JR., ONUFRIEV, A., SIMMER-LING, C., WANG, B., and WOODS, R.J. (2005). The Amberbiomolecular simulation programs. J. Comput. Chem. 26,1668–1688.

CHAKRAVARTHY, U., ADAMIS, A.P., CUNNINGHAM, E.T.,JR., GOLDBAUM, M., GUYER, D.R., KATZ, B., and PATEL,M. (2006). Year 2 efficacy results of 2 randomized controlledclinical trials of pegaptanib for neovascular age-related mac-ular degeneration. Ophthalmology 113, 1508, e1501–e1525.

CHEN, S.J. (2008). RNA folding: conformational statistics, fold-ing kinetics, and ion electrostatics. Annu. Rev. Biophys. 37,197–214.

CHU, T.C., TWU, K.Y., ELLINGTON, A.D., and LEVY, M.(2006). Aptamer mediated siRNA delivery. Nucleic Acids Res.34, e73.

COLOMBATTI, M., GRASSO, S., PORZIA, A., FRACASSO, G.,SCUPOLI, M.T., CINGARLINI, S., POFFE, O., NAIM, H.Y.,HEINE, M., TRIDENTE, G., MAINIERO, F., and RAMARLI,D. (2009). The prostate specific membrane antigen regulatesthe expression of IL-6 and CCL5 in prostate tumour cells byactivating the MAPK pathways. PLoS One 4, e4608.

COSMI, B. (2009). ARC-1779, a PEGylated aptamer antagonist ofvon Willebrand factor for potential use as an anticoagulant orantithrombotic agent. Curr. Opin. Mol. Ther. 11, 322–328.

DARTY, K., DENISE, A., and PONTY, Y. (2009). VARNA: in-teractive drawing and editing of the RNA secondary structure.Bioinformatics 25, 1974–1975.

DASSIE, J.P., LIU, X.Y., THOMAS, G.S., WHITAKER, R.M.,THIEL, K.W., STOCKDALE, K.R., MEYERHOLZ, D.K.,MCCAFFREY, A.P., MCNAMARA, J.O., 2nd, and GIAN-GRANDE, P.H. (2009). Systemic administration of optimizedaptamer-siRNA chimeras promotes regression of PSMA-ex-pressing tumors. Nat. Biotechnol. 27, 839–849.

DAVIS, M.I., BENNETT, M.J., THOMAS, L.M., and BJORK-MAN, P.J. (2005). Crystal structure of prostate-specific mem-brane antigen, a tumor marker and peptidase. Proc. Natl.Acad. Sci. U. S. A. 102, 5981–5986.

DHAR, S., GU, F.X., LANGER, R., FAROKHZAD, O.C., andLIPPARD, S.J. (2008). Targeted delivery of cisplatin to prostatecancer cells by aptamer functionalized Pt(IV) prodrug-PLGA-PEG nanoparticles. Proc. Natl. Acad. Sci. U. S. A. 105, 17356–17361.

DHAR, S., KOLISHETTI, N., LIPPARD, S.J., and FAROKHZAD,O.C. (2011). Targeted delivery of a cisplatin prodrug for saferand more effective prostate cancer therapy in vivo. Proc. Natl.Acad. Sci. U. S. A. 108, 1850–1855.

DYKE, C.K., STEINHUBL, S.R., KLEIMAN, N.S., CANNON,R.O., ABERLE, L.G., LIN, M., MYLES, S.K., MELLONI, C.,HARRINGTON, R.A., ALEXANDER, J.H., BECKER, R.C., andRUSCONI, C.P. (2006). First-in-human experience of an anti-dote-controlled anticoagulant using RNA aptamer technology:a phase 1a pharmacodynamic evaluation of a drug-antidotepair for the controlled regulation of factor IXa activity. Cir-culation 114, 2490–2497.

EIKELBOOM, J.W., ZELENKOFSKE, S.L., and RUSCONI, C.P.(2010). Coagulation factor IXa as a target for treatment andprophylaxis of venous thromboembolism. Arterioscler.Thromb. Vasc. Biol. 30, 382–387.

ELLINGTON, A.D., and SZOSTAK, J.W. (1990). In vitro selectionof RNA molecules that bind specific ligands. Nature 346, 818–822.

FAROKHZAD, O.C., JON, S., KHADEMHOSSEINI, A., TRAN,T.N., LAVAN, D.A., and LANGER, R. (2004). Nanoparticle-aptamer bioconjugates: a new approach for targeting prostatecancer cells. Cancer Res. 64, 7668–7672.

FLOEGE, J., OSTENDORF, T., JANSSEN, U., BURG, M., RA-DEKE, H.H., VARGEESE, C., GILL, S.C., GREEN, L.S., andJANJIC, N. (1999). Novel approach to specific growth factorinhibition in vivo: antagonism of platelet-derived growth factorin glomerulonephritis by aptamers. Am. J. Pathol. 154, 169–179.

GILBERT, J.C., DEFEO-FRAULINI, T., HUTABARAT, R.M.,HORVATH, C.J., MERLINO, P.G., MARSH, H.N., HEALY,J.M., BOUFAKHREDDINE, S., HOLOHAN, T.V., andSCHAUB, R.G. (2007). First-in-human evaluation of anti vonWillebrand factor therapeutic aptamer ARC1779 in healthyvolunteers. Circulation 116, 2678–2686.

GRAGOUDAS, E.S., ADAMIS, A.P., CUNNINGHAM, E.T., JR.,FEINSOD, M., and GUYER, D.R. (2004). Pegaptanib for neo-vascular age-related macular degeneration. N. Engl. J. Med.351, 2805–2816.

GREEN, L., WAUGH, S., BINKLEY, J.P., HOSTOMSKA, Z.,HOSTOMSKY, Z., and TUERK, C. (1995). Comprehensivechemical modification interference and nucleotide substitutionanalysis of an RNA pseudoknot inhibitor to HIV-1 reversetranscriptase. J. Mol. Biol. 247, 60–68.

312 ROCKEY ET AL.

GU, F., ZHANG, L., TEPLY, B.A., MANN, N., WANG, A.,RADOVIC-MORENO, A.F., LANGER, R., and FAROKHZAD,O.C. (2008). Precise engineering of targeted nanoparticles byusing self-assembled biointegrated block copolymers. Proc.Natl. Acad. Sci. U. S. A. 105, 2586–2591.

HEALY, J.M., LEWIS, S.D., KURZ, M., BOOMER, R.M.,THOMPSON, K.M., WILSON, C., and MCCAULEY, T.G.(2004). Pharmacokinetics and biodistribution of novel aptamercompositions. Pharm. Res. 21, 2234–2246.

HERNANDEZ, F.J., KALRA, N., WENGEL, J., and VESTER,B. (2009). Aptamers as a model for functional evaluation of LNAand 2’-amino LNA. Bioorg. Med. Chem. Lett. 19, 6585–6587.

HICKE, B.J., STEPHENS, A.W., GOULD, T., CHANG, Y.F.,LYNOTT, C.K., HEIL, J., BORKOWSKI, S., HILGER, C.S.,COOK, G., WARREN, S., and SCHMIDT, P.G. (2006). Tumortargeting by an aptamer. J. Nucl. Med. 47, 668–678.

HUANG, Y., ECKSTEIN, F., PADILLA, R., and SOUSA, R.(1997). Mechanism of ribose 2’-group discrimination by anRNA polymerase. Biochemistry 36, 8231–8242.

HUANG, S.Y., and ZOU, X. (2008). An iterative knowledge-based scoring function for protein-protein recognition. Pro-teins 72, 557–579.

HUANG, S.Y., and ZOU, X. (2010). MDockPP: A hierarchicalapproach for protein-protein docking and its application toCAPRI rounds 15–19. Proteins 78, 3096–3103.

JELLINEK, D., GREEN, L.S., BELL, C., LYNOTT, C.K., GILL, N.,VARGEESE, C., KIRSCHENHEUTER, G., MCGEE, D.P.,ABESINGHE, P., PIEKEN, W.A., et al. (1995). Potent 2’-amino-2’-deoxypyrimidine RNA inhibitors of basic fibroblast growthfactor. Biochemistry 34, 11363–11372.

KATILIUS, E., FLORES, C., and WOODBURY, N.W. (2007).Exploring the sequence space of a DNA aptamer using mi-croarrays. Nucleic Acids Res. 35, 7626–7635.

KAWAGUCHI, T., ASAKAWA, H., TASHIRO, Y., JUNI, K., andSUEISHI, T. (1995). Stability, specific binding activity, andplasma concentration in mice of an oligodeoxynucleotidemodified at 5’-terminal with poly(ethylene glycol). Biol.Pharm. Bull. 18, 474–476.

KEEFE, A.D., PAI, S., and ELLINGTON, A. (2010). Aptamers astherapeutics. Nat. Rev. Drug Discov. 9, 537–550.

KULARATNE, S.A., WANG, K., SANTHAPURAM, H.K., andLOW, P.S. (2009). Prostate-specific membrane antigen targetedimaging and therapy of prostate cancer using a PSMA inhib-itor as a homing ligand. Mol. Pharm. 6, 780–789.

LAPIDUS, R.G., TIFFANY, C.W., ISAACS, J.T., and SLUSHER,B.S. (2000). Prostate-specific membrane antigen (PSMA) en-zyme activity is elevated in prostate cancer cells. Prostate 45,350–354.

LIN, Y., QIU, Q., GILL, S.C., and JAYASENA, S.D. (1994).Modified RNA sequence pools for in vitro selection. NucleicAcids Res. 22, 5229–5234.

LUPOLD, S.E., HICKE, B.J., LIN, Y., and COFFEY, D.S. (2002).Identification and characterization of nuclease-stabilized RNAmolecules that bind human prostate cancer cells via theprostate-specific membrane antigen. Cancer Res. 62, 4029–4033.

MATHEWS, D.H. (2006). RNA secondary structure analysisusing RNA structure. Curr. Protoc. Bioinform. Chapter 12,Unit 12.6.

MATHEWS, D.H., TURNER, D.H., and ZUKER, M. (2007). RNAsecondary structure prediction. Curr. Protoc. Nucleic AcidChem. Chapter 11, Unit 11.2.

MCNAMARA, J.O., 2nd, ANDRECHEK, E.R., WANG, Y.,VILES, K.D., REMPEL, R.E., GILBOA, E., SULLENGER, B.A.,

and GIANGRANDE, P.H. (2006). Cell type-specific delivery ofsiRNAs with aptamer-siRNA chimeras. Nat. Biotechnol. 24,1005–1015.

MCNAMARA, J.O., KOLONIAS, D., PASTOR, F., MITTLER,R.S., CHEN, L., GIANGRANDE, P.H., SULLENGER, B., andGILBOA, E. (2008). Multivalent 4-1BB binding aptamerscostimulate CD8 + T cells and inhibit tumor growth in mice. J.Clin. Invest. 118, 376–386.

MONGELARD, F., and BOUVET, P. (2010). AS-1411, a guano-sine-rich oligonucleotide aptamer targeting nucleolin for thepotential treatment of cancer, including acute myeloid leuke-mia. Curr. Opin. Mol. Ther. 12, 107–114.

NEFF, C.P., ZHOU, J., REMLING, L., KURUVILLA, J., ZHANG,J., LI, H., SMITH, D.D., SWIDERSKI, P., ROSSI, J.J., and AK-KINA, R. (2011). An aptamer-siRNA chimera suppresses HIV-1 viral loads and protects from helper CD4( + ) T cell decline inhumanized mice. Sci. Transl. Med. 3, 66ra6.

NG, E.W., and ADAMIS, A.P. (2006). Anti-VEGF aptamer (pe-gaptanib) therapy for ocular vascular diseases. Ann. N.Y.Acad. Sci. 1082, 151–171.

PADILLA, R., and SOUSA, R. (1999). Efficient synthesis of nu-cleic acids heavily modified with non-canonical ribose 2’-groups using a mutantT7 RNA polymerase (RNAP). NucleicAcids Res. 27, 1561–1563.

PASTOR, F., KOLONIAS, D., GIANGRANDE, P.H., and GILBOA,E. (2010). Induction of tumour immunity by targeted inhibition ofnonsense-mediated mRNA decay. Nature 465, 227–230.

REESE, C.B. (2005). Oligo- and poly-nucleotides: 50 years ofchemical synthesis. Org. Biomol. Chem. 3, 3851–3868.

RUCKMAN, J., GREEN, L.S., BEESON, J., WAUGH, S., GILL-ETTE, W.L., HENNINGER, D.D., CLAESSON-WELSH, L.,and JANJIC, N. (1998). 2’-Fluoropyrimidine RNA-based ap-tamers to the 165-amino acid form of vascular endothelialgrowth factor (VEGF165). Inhibition of receptor binding andVEGF-induced vascular permeability through interactions re-quiring the exon 7-encoded domain. J. Biol. Chem. 273, 20556–20567.

RUSCONI, C.P., SCARDINO, E., LAYZER, J., PITOC, G.A.,ORTEL, T.L., MONROE, D., and SULLENGER, B.A. (2002).RNA aptamers as reversible antagonists of coagulation factorIXa. Nature 419, 90–94.

SASSANFAR, M., and SZOSTAK, J.W. (1993). An RNA motifthat binds ATP. Nature 364, 550–553.

SERRA, M.J., and TURNER, D.H. (1995). Predicting thermody-namic properties of RNA. Methods Enzymol. 259, 242–261.

SILVER, D.A., PELLICER, I., FAIR, W.R., HESTON, W.D., andCORDON-CARDO, C. (1997). Prostate-specific membraneantigen expression in normal and malignant human tissues.Clin. Cancer Res. 3, 81–85.

SOONTORNWORAJIT, B., ZHOU, J., SNIPES, M.P., BATTIG,M.R., and WANG, Y. (2011). Affinity hydrogels for controlledprotein release using nucleic acid aptamers and complemen-tary oligonucleotides. Biomaterials 32, 6839–6849.

THIEL, K.W., and GIANGRANDE, P.H. (2009). Therapeuticapplications of DNA and RNA aptamers. Oligonucleotides 19,209–222.

WATSON, S.R., CHANG, Y.F., O’CONNELL, D., WEIGAND,L., RINGQUIST, S., and PARMA, D.H. (2000). Anti-L-selectinaptamers: binding characteristics, pharmacokinetic parame-ters, and activity against an intravascular target in vivo. An-tisense Nucleic Acid Drug Dev. 10, 63–75.

WONG, I., and LOHMAN, T.M. (1993). A double-filter methodfor nitrocellulose-filter binding: application to protein-nucleicacid interactions. Proc. Natl. Acad. Sci. U. S. A. 90, 5428–5432.

COMPUTER-GUIDED OPTIMIZATION OF AN RNA APTAMER 313

XIAO, Z., JIANG, X., BECKETT, M.L., and WRIGHT, G.L., JR.(2000). Generation of a baculovirus recombinant prostate-specific membrane antigen and its use in the development of anovel protein biochip quantitative immunoassay. ProteinExpr. Purif. 19, 12–21.

YAO, V., BERKMAN, C.E., CHOI, J.K., O’KEEFE, D.S., andBACICH, D.J. (2010). Expression of prostate-specific mem-brane antigen (PSMA), increases cell folate uptake and pro-liferation and suggests a novel role for PSMA in the uptake ofthe non-polyglutamated folate, folic acid. Prostate 70, 305–316.

ZHOU, J., LI, H., LI, S., ZAIA, J., and ROSSI, J.J. (2008). Noveldual inhibitory function aptamer-siRNA delivery system forHIV-1 therapy. Mol. Ther. 16, 1481–1489.

ZHOU, J., SOONTORNWORAJIT, B., SNIPES, M.P., andWANG, Y. (2011). Structural prediction and binding analysisof hybridized aptamers. J. Mol. Recognit. 24, 119–126.

ZHOU, J., SWIDERSKI, P., LI, H., ZHANG, J., NEFF, C.P.,AKKINA, R., and ROSSI, J.J. (2009). Selection, characterization

and application of new RNA HIV gp 120 aptamers for faciledelivery of Dicer substrate siRNAs into HIV infected cells.Nucleic Acids Res. 37, 3094–3109.

Address correspondence to:Dr. Paloma H. Giangrande

Department of Internal MedicineUniversity of Iowa285 Newton Road

5202 MERFIowa City, IA 52242

E-mail: paloma-giangrande@uiowa.edu

Received for publication July 7, 2011; accepted after revisionAugust 18, 2011.

314 ROCKEY ET AL.